Apparatus and method of recognizing user postures

Patent metadata

Field	Value
Publication number	US-10297041-B2
Application number	US-201715484258-A
Country	US
Kind code	B2
Filing date	Apr 11, 2017
Priority date	Apr 11, 2016
Publication date	May 21, 2019
Grant date	May 21, 2019

Abstract

Provided is an apparatus for recognizing user postures. The apparatus recognizes detailed postures such as a stand posture, a bending forward posture, a bending knees posture, a tilt right posture, and a tilt left posture, based on a variation between body information at a previous time and body information at a current time.

First claim

What is claimed is:

1. A method of recognizing user postures, the method performed by an apparatus having an image processor and a three-dimensional (3D) camera, the method comprising:
extracting, from an image frame input through the 3D camera, body information including a two-dimensional (2D) coordinate value “H” representing a head position of a user, a three-dimensional (3D) coordinate value “M” representing a body middle position of the user, a 2D coordinate value “C” representing a center position between a left ankle position and a right ankle position of the user, a distance value “HM” from the head position of the user to the body middle position of the user, a distance value “HC” from the head position of the user to the center position between the left ankle position and the right ankle position of the user, and a distance value “MC” from the body middle position of the user to the center position “C” between the left ankle position “L” and the right ankle position “R” of the user; and
recognizing a posture of the user, based on the body information sequentially extracted in units of image frames from the 3D camera.

2. The method of claim 1 , wherein the recognizing of the posture comprises:
(A) determining whether a value obtained by summating a first HC input at a n th time and a first reference value is less than a second HC input at a n−1 th time;
(B) determining whether a value obtained by summating a z coordinate value of a first M input at the n−1 th time and a second reference value is less than a z coordinate value of a second M input at the n th time;
(C) determining whether a value obtained by summating a first HM input at the n th time and a third reference value is less than a second HM input at the n−1 th time;
(D) determining whether an absolute value of a difference value between a first MC input at the n−1 th time and a second MC input at the n th time is equal to or less than a fourth reference value; and
(E) when steps (A), (B), (C), and (D) are all satisfied, recognizing the posture of user as a bending forward posture.

3. The method of claim 1 , wherein the recognizing of the posture comprises:
(A) determining whether a value obtained by summating a first HC input at a n th time and a first reference value is less than a second HC input at a n−1 th time;
(B) determining whether a value obtained by summating a z coordinate value of a first M input at the n th time and a second reference value is less than a z coordinate value of M input at the n−1 th time;
(C) determining whether an absolute value of a difference value between a first HM input at the n−1 th time and HM input at the n th time is equal to or less than a third reference value;
(D) determining whether a value obtained by summating a first MC input at the n th time and a fourth reference value is less than a second MC input at the n−1 th time; and
(E) when steps (A), (B), (C), and (D) are all satisfied, recognizing the posture of user as a bending knees posture.

4. The method of claim 1 , wherein the recognizing of the posture comprises:
(A) determining whether a difference between an x coordinate value of an H input at a n th and an x coordinate value of a C input at the n th time is a positive number and is more than a fifth reference value;
(B) determining whether a difference between an x coordinate value of an M input at the n th time and the x coordinate value of the C input at the n th time is a positive number and is more than a sixth reference value; and
(C) when steps (A) and (B) are all satisfied, recognizing the posture of user as a tilt right posture.

5. The method of claim 1 , wherein the recognizing of the posture comprises:
(A) determining whether a difference between an x coordinate value of a C input at a n th time and an x coordinate value of H input at the n th time is a positive number and is more than a fifth reference value;
(B) determining whether a difference between the x coordinate value of the C input at the n th time and an x coordinate value of an M input at the n th time is a positive number and is more than a sixth reference value; and
(C) when steps (A) and (B) are all satisfied, recognizing the posture of user as a tilt left posture.

6. An apparatus for recognizing user postures, the apparatus comprising:
a three-dimensional (3D) camera configured to take an image of a user; and
an image processor configured to extract body information from the image, the body information including a two-dimensional (2D) coordinate value “H” representing a head position of a user, a 3D coordinate value “M” representing a body middle position of the user, a 2D coordinate value “C” representing a center position between a left ankle position and a right ankle position of the user, a distance value “HM” from the head position of the user to the body middle position of the user, a distance value “HC” from the head position of the user to the center position between the left ankle position and the right ankle position of the user, and a distance value “MC” from the body middle position of the user to the center position “C” between the left ankle position “L” and the right ankle position “R” of the user; and
a posture recognizer configured to sequentially receive body information in units of image frames from the image processor and recognize a posture of the user, based on the body information.

7. The apparatus of claim 6 , wherein
the posture recognizer recognizes one of a bending forward posture, and a bending knees posture, based on a variation of the HC, a variation of the M, a variation of the HM, and a variation of the MC, and
the posture recognizer recognizes one of a tilt right posture and a tilt left posture, based on a difference value between an x coordinate value of the H and an x coordinate value of the C and a difference value between an x coordinate value of the M and the x coordinate value of the C.

Assignees

• Korea Electronics Technology

Inventors

• Ahn Yang Keun
• Jung Kwang Mo
• Kim Chan Gyu

Classifications

• G06V40/103 Primary

  Static body considered as a whole, e.g. static pedestrian or occupant recognition · CPC title

• G06T7/73 Primary

  using feature-based methods · CPC title

• G06V10/422

  for representing the structure of the pattern or shape of an object therefor · CPC title

• G06V10/757

  Matching configurations of points or features · CPC title

• G06T2207/10012

  Stereo images · CPC title